Selection between Granger causality in level, Granger causality/Block exogenity and Granger causality test at first differences?
1) If time series variables are non-stationary at level data, then as per my knowledge we can not use Granger casuality test because it breaks the assumption underlying the test.
So should I use Granger casuality/Block exogenity test ( avaliable in EViews software) or Granger causality test at first differences?
2) Each of the above test shows which type of causality- Either it is short-run causality or long-run causality?
3) How to determine the lag length?
Thank you.